Remove qas input wherever possible. Comment out NQ inference, since i…

…t is not working (#681)
deepset-ai · Jan 8, 2021 · d4f4508 · d4f4508
1 parent 42af265
commit d4f4508
Show file tree

Hide file tree

Showing 6 changed files with 22 additions and 55 deletions.
diff --git a/examples/inferencer_multiprocessing.py b/examples/inferencer_multiprocessing.py
diff --git a/examples/natural_questions.py b/examples/natural_questions.py
@@ -122,21 +122,23 @@ def question_answering():
     model.save(save_dir)
     processor.save(save_dir)
 
-    # 9. Since training on the whole NQ corpus requires substantial compute resources we trained and uploaded a model on s3
-    fetch_archive_from_http("https://s3.eu-central-1.amazonaws.com/deepset.ai-farm-qa/models/roberta-base-squad2-nq.zip", output_dir="../saved_models/farm")
-    QA_input = [
-        {
-            "qas": ["Did GameTrailers rated Twilight Princess as one of the best games ever created?"],
-            "context":  "Twilight Princess was released to universal critical acclaim and commercial success. It received perfect scores from major publications such as 1UP.com, Computer and Video Games, Electronic Gaming Monthly, Game Informer, GamesRadar, and GameSpy. On the review aggregators GameRankings and Metacritic, Twilight Princess has average scores of 95% and 95 for the Wii version and scores of 95% and 96 for the GameCube version. GameTrailers in their review called it one of the greatest games ever created."
-        }
-    ]
-
-    model = QAInferencer.load(model_name_or_path="../saved_models/farm/roberta-base-squad2-nq", batch_size=batch_size, gpu=True)
-    result = model.inference_from_dicts(dicts=QA_input, return_json=False) # result is a list of QAPred objects
 
-    print(f"\nQuestion: Did GameTrailers rated Twilight Princess as one of the best games ever created?"
-          f"\nAnswer from model: {result[0].prediction[0].answer}")
-    model.close_multiprocessing_pool()
+    # TODO make inferencing work again with trained and saved model (currently the inferencer loads a fast Robertatokenizer, which is not supported by NQ)
+    # # 9. Since training on the whole NQ corpus requires substantial compute resources we trained and uploaded a model on s3
+    # fetch_archive_from_http("https://s3.eu-central-1.amazonaws.com/deepset.ai-farm-qa/models/roberta-base-squad2-nq.zip", output_dir="../saved_models/farm")
+    # QA_input = [
+    #     {
+    #         "qas": ["Did GameTrailers rated Twilight Princess as one of the best games ever created?"],
+    #         "context":  "Twilight Princess was released to universal critical acclaim and commercial success. It received perfect scores from major publications such as 1UP.com, Computer and Video Games, Electronic Gaming Monthly, Game Informer, GamesRadar, and GameSpy. On the review aggregators GameRankings and Metacritic, Twilight Princess has average scores of 95% and 95 for the Wii version and scores of 95% and 96 for the GameCube version. GameTrailers in their review called it one of the greatest games ever created."
+    #     }
+    # ]
+    #
+    # model = QAInferencer.load(model_name_or_path="../saved_models/farm/roberta-base-squad2-nq", batch_size=batch_size, gpu=True)
+    # result = model.inference_from_dicts(dicts=QA_input, return_json=False) # result is a list of QAPred objects
+    #
+    # print(f"\nQuestion: Did GameTrailers rated Twilight Princess as one of the best games ever created?"
+    #       f"\nAnswer from model: {result[0].prediction[0].answer}")
+    # model.close_multiprocessing_pool()
 
 if __name__ == "__main__":
     question_answering()

diff --git a/examples/onnx_question_answering.py b/examples/onnx_question_answering.py
@@ -20,8 +20,8 @@ def onnx_runtime_example():
 
     qa_input = [
         {
-            "qas": ["Who counted the game among the best ever made?"],
-            "context": "Twilight Princess was released to universal critical acclaim and commercial success. "
+            "questions": ["Who counted the game among the best ever made?"],
+            "text": "Twilight Princess was released to universal critical acclaim and commercial success. "
             "It received perfect scores from major publications such as 1UP.com, Computer and Video Games, "
             "Electronic Gaming Monthly, Game Informer, GamesRadar, and GameSpy. On the review aggregators "
             "GameRankings and Metacritic, Twilight Princess has average scores of 95% and 95 for the Wii "

diff --git a/examples/question_answering.py b/examples/question_answering.py
@@ -106,8 +106,8 @@ def question_answering():
     # 9. Load it & harvest your fruits (Inference)
     QA_input = [
             {
-                "qas": ["Who counted the game among the best ever made?"],
-                "context":  "Twilight Princess was released to universal critical acclaim and commercial success. It received perfect scores from major publications such as 1UP.com, Computer and Video Games, Electronic Gaming Monthly, Game Informer, GamesRadar, and GameSpy. On the review aggregators GameRankings and Metacritic, Twilight Princess has average scores of 95% and 95 for the Wii version and scores of 95% and 96 for the GameCube version. GameTrailers in their review called it one of the greatest games ever created."
+                "questions": ["Who counted the game among the best ever made?"],
+                "text":  "Twilight Princess was released to universal critical acclaim and commercial success. It received perfect scores from major publications such as 1UP.com, Computer and Video Games, Electronic Gaming Monthly, Game Informer, GamesRadar, and GameSpy. On the review aggregators GameRankings and Metacritic, Twilight Princess has average scores of 95% and 95 for the Wii version and scores of 95% and 96 for the GameCube version. GameTrailers in their review called it one of the greatest games ever created."
             }]
 
     model = QAInferencer.load(save_dir, batch_size=40, gpu=True)

diff --git a/examples/streaming_inference.py b/examples/streaming_inference.py
@@ -33,8 +33,8 @@ def sample_dicts_generator():
     :rtype: iter
     """
     qa_input = {
-        "qas": ["Who counted the game among the best ever made?"],
-        "context": "Twilight Princess was released to universal critical acclaim and commercial success. "
+        "questions": ["Who counted the game among the best ever made?"],
+        "text": "Twilight Princess was released to universal critical acclaim and commercial success. "
                    "It received perfect scores from major publications such as 1UP.com, Computer and Video Games, "
                    "Electronic Gaming Monthly, Game Informer, GamesRadar, and GameSpy. On the review aggregators "
                    "GameRankings and Metacritic, Twilight Princess has average scores of 95% and 95 for the Wii "

diff --git a/farm/infer.py b/farm/infer.py
@@ -389,7 +389,6 @@ def inference_from_dicts(
         Runs down-stream inference on samples created from input dictionaries.
         The format of the input `dicts` depends on the task:
 
-        * QA (SQuAD style):    [{"qas": ["What is X?"], "context":  "Some context containing the answer"}] (Deprecated)
         * QA (FARM style): [{"questions": ["What is X?"], "text":  "Some context containing the answer"}]
         * Classification / NER / embeddings: [{"text": "Some input text"}]